Eugene  

Sometimes (near every year), I touch the accessible for me LLMs to see how they able to perform their main task — operate with human-written text input to give me some valid output from their giant databases.

And looks like there are not a big progress  I queried two LLMs: Claude and the ChatGPT 4.x — trying to find one indie game. All that I remember: this game was made by a single Russian developer, the main plot depicts a retro-futuristic conflict somewhere in the Middle Asia, between a fiction empire, which looks like a Russian Empire and some powerful traitor, and gameplay contains flying diesel-punk ships, filled with machine-guns, rockets, atomic bombs, etc. And the game name starts from "H" letter.

Claude gave me a lot of irrelevant game titles. And a ChatGPT gave me …. "an error occurred (code 500)  

As a result, after spending near 10 minutes in the LLM chats I remembered the game name by myself — it was a "HighFleet". Another flawless victory of natural intelligence, lol  

#llm #claude #chatgpt #highfleet

Nov 21 at 10:05pmWeb
Show threadLFA 🇮🇨Nov 22

@evgandr Techbross mode on:
No, no, no, no way. What you think is your own intelligence that made you remember the name was in fact a new feature of those superior artificial intelligences. We call it telepathic answer and basically it'll send you a telepathic answer while writing nonsense on the app in order to make your brain more receptive to the telepathic message. Currently we're testing it with selected premium people before releasing it to the broad public.

😂